Therapeutic Dose: Extending QbD to Stability Testing

April 8, 2008
The concept of a process signature can easily be extended to stability testing.

These are exciting times for any physical or analytical chemists working in the pharmaceutical industry, as the concepts of process analytical technologies (PAT) and Quality by Design (QbD) are applied to pharmaceutical manufacturing. The “design” part of QbD is being extended to control not only the ongoing process, but the raw materials. Practitioners suggest that we look as closely at excipients as we currently do at active pharmaceutical ingredients (APIs).

As a result, more companies are staring to measure things such as compressabilty, porosity, morphology, mean particle size and distribution, degree of crystallinity, and other physical parameters. We have developed the concept of “process signature” to assure that all the final parameters of a product will meet specifications such as content uniformity and dissolution release times. This signature, often a spectroscopic signature (spectrum), is interpreted via chemometrics.

Prior batches/samples of good, marginal, and failing dosages are used to build the chemometric equation, which is then used to determine “goodness.” Many effects to the spectrum are physical as well as chemical and are not isolated. It has been suggested that this process signature be used to detect counterfeit materials. The software may not specifically “know” what’s different, but the signature will be clearly seen as different. Expanding this concept to stability testing does not require a large leap of faith.

Currently, even in a PAT setting, we only assay up to ten tablets or capsules for content uniformity and six for dissolution profile. It does not seem to make sense that we test thousands or even tens of thousands of dosage forms during a run to assure conformity, then fall back to a tiny fraction of a percent for stability testing. A number of years ago, I did some work investigating why pellication can occur in capsules.

Named for the large-billed bird, this occurs when a bulge or aneurism forms on a capsule instead of it merely rupturing during dissolution. The problem was caused by cross-linking of the gelatin over time. It didn’t occur on all the capsules, so chance played a major part in whether any batch passed or failed. If, say, six of 100 capsules had crosslinking, whether the lot passed or failed depended on which capsules were drawn from the bottle at that point in time.

That is, the same composition of product could pass or fail, based on chance! We used NIR to inspect an entire bottle of capsules over time, to look for changes. If a spectral change occurred, similar to earlier deliberate cross-linking experiments, we assayed some of the changed capsules via dissolution, as well as some of the unchanged ones.

Invariably, the changed capsules displayed some amount of pellication. If these types of methods were extended to both tablets and capsules, stability could become both more predictive and an ally in QbD. What if we placed a number of bottles of a dosage form on stability, just as we now do, but instead of testing a small number of tablets or capsules, we examined the whole bottle? If we “looked” at all 50 or 100 dosage units, we could begin to see changes with time.

However, instead of relying on dumb luck, we could test both changed and unchanged units. We may find, as an example, after six months, four tablets from lot “A” and eight tablets from lot “B” were changed. If we tested both types, we would be covering our compendial duties, plus, as in most PAT projects, generating many, many more assay results. This approach would go a long way in assuring the company and Agency that the most stabile product possible is on the pharmacists’ shelves. Perhaps at one year, the number might be six for “A” and 15 for “B,” and so forth.

Over time, we would have to develop new pass/fail criteria, but that is true for PAT as well. The really exciting part of this is that we will be able, over time, to correlate the rate of change (to “different”) for each lot of each material. These rates can then be correlated to changes in design space. Adding these data to our DoE (design of experiments), we will now see longterm effects for different process parameters.

The end effect may well be a tighter design space designed for the longest product shelf-life as well as for release of the product. Then we would truly have Quality by Design.

About the Author

Emil W. Ciurczak | Contributing Editor